Commit
•
acfda83
1
Parent(s):
5273d7c
Model save
Browse files- README.md +77 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +826 -0
README.md
ADDED
|
@@ -0,0 +1,77 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- dpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-7b-uf-dpo-2e
|
| 11 |
+
results: []
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
+
|
| 17 |
+
# zephyr-7b-uf-dpo-2e
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on an unknown dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.5019
|
| 22 |
+
- Rewards/chosen: -1.5195
|
| 23 |
+
- Rewards/rejected: -2.7388
|
| 24 |
+
- Rewards/accuracies: 0.7695
|
| 25 |
+
- Rewards/margins: 1.2193
|
| 26 |
+
- Logps/rejected: -536.5389
|
| 27 |
+
- Logps/chosen: -414.5813
|
| 28 |
+
- Logits/rejected: 2.1782
|
| 29 |
+
- Logits/chosen: 1.4005
|
| 30 |
+
|
| 31 |
+
## Model description
|
| 32 |
+
|
| 33 |
+
More information needed
|
| 34 |
+
|
| 35 |
+
## Intended uses & limitations
|
| 36 |
+
|
| 37 |
+
More information needed
|
| 38 |
+
|
| 39 |
+
## Training and evaluation data
|
| 40 |
+
|
| 41 |
+
More information needed
|
| 42 |
+
|
| 43 |
+
## Training procedure
|
| 44 |
+
|
| 45 |
+
### Training hyperparameters
|
| 46 |
+
|
| 47 |
+
The following hyperparameters were used during training:
|
| 48 |
+
- learning_rate: 5e-07
|
| 49 |
+
- train_batch_size: 8
|
| 50 |
+
- eval_batch_size: 8
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- num_devices: 8
|
| 54 |
+
- gradient_accumulation_steps: 4
|
| 55 |
+
- total_train_batch_size: 256
|
| 56 |
+
- total_eval_batch_size: 64
|
| 57 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
+
- lr_scheduler_type: cosine
|
| 59 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
+
- num_epochs: 2
|
| 61 |
+
|
| 62 |
+
### Training results
|
| 63 |
+
|
| 64 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 65 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 66 |
+
| 0.5692 | 0.4184 | 100 | 0.5671 | -0.3691 | -0.9295 | 0.7344 | 0.5604 | -355.6129 | -299.5389 | -1.1363 | -1.3128 |
|
| 67 |
+
| 0.5056 | 0.8368 | 200 | 0.5104 | -0.9550 | -1.8507 | 0.7852 | 0.8957 | -447.7286 | -358.1263 | 1.4954 | 0.9604 |
|
| 68 |
+
| 0.3744 | 1.2552 | 300 | 0.5120 | -1.3455 | -2.5618 | 0.7617 | 1.2163 | -518.8406 | -397.1791 | 1.8527 | 0.9871 |
|
| 69 |
+
| 0.351 | 1.6736 | 400 | 0.5019 | -1.5195 | -2.7388 | 0.7695 | 1.2193 | -536.5389 | -414.5813 | 2.1782 | 1.4005 |
|
| 70 |
+
|
| 71 |
+
|
| 72 |
+
### Framework versions
|
| 73 |
+
|
| 74 |
+
- Transformers 4.44.1
|
| 75 |
+
- Pytorch 2.1.2+cu121
|
| 76 |
+
- Datasets 2.21.0
|
| 77 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.47010813497599196,
|
| 5 |
+
"train_runtime": 14181.5338,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 8.622,
|
| 8 |
+
"train_steps_per_second": 0.034
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"transformers_version": "4.44.1"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.47010813497599196,
|
| 5 |
+
"train_runtime": 14181.5338,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 8.622,
|
| 8 |
+
"train_steps_per_second": 0.034
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,826 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 2.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 478,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0041841004184100415,
|
| 13 |
+
"grad_norm": 7.099414444882459,
|
| 14 |
+
"learning_rate": 1.0416666666666666e-08,
|
| 15 |
+
"logits/chosen": -2.6615781784057617,
|
| 16 |
+
"logits/rejected": -2.691082239151001,
|
| 17 |
+
"logps/chosen": -298.690673828125,
|
| 18 |
+
"logps/rejected": -323.7967834472656,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.0,
|
| 21 |
+
"rewards/chosen": 0.0,
|
| 22 |
+
"rewards/margins": 0.0,
|
| 23 |
+
"rewards/rejected": 0.0,
|
| 24 |
+
"step": 1
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.04184100418410042,
|
| 28 |
+
"grad_norm": 6.23352657034775,
|
| 29 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 30 |
+
"logits/chosen": -2.628962278366089,
|
| 31 |
+
"logits/rejected": -2.581556797027588,
|
| 32 |
+
"logps/chosen": -270.5624694824219,
|
| 33 |
+
"logps/rejected": -274.7192077636719,
|
| 34 |
+
"loss": 0.6931,
|
| 35 |
+
"rewards/accuracies": 0.4375,
|
| 36 |
+
"rewards/chosen": 0.00030235512531362474,
|
| 37 |
+
"rewards/margins": 0.00018380279652774334,
|
| 38 |
+
"rewards/rejected": 0.00011855230695800856,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.08368200836820083,
|
| 43 |
+
"grad_norm": 6.403910934005757,
|
| 44 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 45 |
+
"logits/chosen": -2.6337690353393555,
|
| 46 |
+
"logits/rejected": -2.5815837383270264,
|
| 47 |
+
"logps/chosen": -283.6520080566406,
|
| 48 |
+
"logps/rejected": -254.223876953125,
|
| 49 |
+
"loss": 0.692,
|
| 50 |
+
"rewards/accuracies": 0.606249988079071,
|
| 51 |
+
"rewards/chosen": 0.0022390992380678654,
|
| 52 |
+
"rewards/margins": 0.002958392258733511,
|
| 53 |
+
"rewards/rejected": -0.000719292729627341,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.12552301255230125,
|
| 58 |
+
"grad_norm": 6.382530025546531,
|
| 59 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 60 |
+
"logits/chosen": -2.596775531768799,
|
| 61 |
+
"logits/rejected": -2.5497207641601562,
|
| 62 |
+
"logps/chosen": -273.9563293457031,
|
| 63 |
+
"logps/rejected": -294.90093994140625,
|
| 64 |
+
"loss": 0.6863,
|
| 65 |
+
"rewards/accuracies": 0.640625,
|
| 66 |
+
"rewards/chosen": 0.01745353825390339,
|
| 67 |
+
"rewards/margins": 0.0119100296869874,
|
| 68 |
+
"rewards/rejected": 0.005543507635593414,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.16736401673640167,
|
| 73 |
+
"grad_norm": 7.236510898066947,
|
| 74 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 75 |
+
"logits/chosen": -2.595787525177002,
|
| 76 |
+
"logits/rejected": -2.56512188911438,
|
| 77 |
+
"logps/chosen": -289.8186340332031,
|
| 78 |
+
"logps/rejected": -260.59027099609375,
|
| 79 |
+
"loss": 0.6711,
|
| 80 |
+
"rewards/accuracies": 0.699999988079071,
|
| 81 |
+
"rewards/chosen": 0.044044919312000275,
|
| 82 |
+
"rewards/margins": 0.045689694583415985,
|
| 83 |
+
"rewards/rejected": -0.0016447717789560556,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.20920502092050208,
|
| 88 |
+
"grad_norm": 8.532614175897798,
|
| 89 |
+
"learning_rate": 4.999733114418725e-07,
|
| 90 |
+
"logits/chosen": -2.5427632331848145,
|
| 91 |
+
"logits/rejected": -2.5035042762756348,
|
| 92 |
+
"logps/chosen": -262.0881042480469,
|
| 93 |
+
"logps/rejected": -270.3690185546875,
|
| 94 |
+
"loss": 0.6456,
|
| 95 |
+
"rewards/accuracies": 0.699999988079071,
|
| 96 |
+
"rewards/chosen": -0.023291397839784622,
|
| 97 |
+
"rewards/margins": 0.13293734192848206,
|
| 98 |
+
"rewards/rejected": -0.15622875094413757,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.2510460251046025,
|
| 103 |
+
"grad_norm": 10.272678361666578,
|
| 104 |
+
"learning_rate": 4.990398100856366e-07,
|
| 105 |
+
"logits/chosen": -2.531106472015381,
|
| 106 |
+
"logits/rejected": -2.470402240753174,
|
| 107 |
+
"logps/chosen": -302.3480224609375,
|
| 108 |
+
"logps/rejected": -291.5530700683594,
|
| 109 |
+
"loss": 0.6203,
|
| 110 |
+
"rewards/accuracies": 0.7093750238418579,
|
| 111 |
+
"rewards/chosen": -0.09494151920080185,
|
| 112 |
+
"rewards/margins": 0.23186612129211426,
|
| 113 |
+
"rewards/rejected": -0.3268076777458191,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.2928870292887029,
|
| 118 |
+
"grad_norm": 14.725483063714895,
|
| 119 |
+
"learning_rate": 4.967775735898179e-07,
|
| 120 |
+
"logits/chosen": -2.4730429649353027,
|
| 121 |
+
"logits/rejected": -2.4408392906188965,
|
| 122 |
+
"logps/chosen": -316.3986511230469,
|
| 123 |
+
"logps/rejected": -333.5642395019531,
|
| 124 |
+
"loss": 0.5917,
|
| 125 |
+
"rewards/accuracies": 0.706250011920929,
|
| 126 |
+
"rewards/chosen": -0.1742347776889801,
|
| 127 |
+
"rewards/margins": 0.31590718030929565,
|
| 128 |
+
"rewards/rejected": -0.49014195799827576,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.33472803347280333,
|
| 133 |
+
"grad_norm": 10.498750585183611,
|
| 134 |
+
"learning_rate": 4.931986719649298e-07,
|
| 135 |
+
"logits/chosen": -2.521245002746582,
|
| 136 |
+
"logits/rejected": -2.4482836723327637,
|
| 137 |
+
"logps/chosen": -308.7649841308594,
|
| 138 |
+
"logps/rejected": -326.46234130859375,
|
| 139 |
+
"loss": 0.6031,
|
| 140 |
+
"rewards/accuracies": 0.6968749761581421,
|
| 141 |
+
"rewards/chosen": -0.30625176429748535,
|
| 142 |
+
"rewards/margins": 0.29028043150901794,
|
| 143 |
+
"rewards/rejected": -0.5965322256088257,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.37656903765690375,
|
| 148 |
+
"grad_norm": 12.712503521629706,
|
| 149 |
+
"learning_rate": 4.883222001996351e-07,
|
| 150 |
+
"logits/chosen": -2.414280414581299,
|
| 151 |
+
"logits/rejected": -2.352023124694824,
|
| 152 |
+
"logps/chosen": -291.20147705078125,
|
| 153 |
+
"logps/rejected": -300.1272888183594,
|
| 154 |
+
"loss": 0.5797,
|
| 155 |
+
"rewards/accuracies": 0.6875,
|
| 156 |
+
"rewards/chosen": -0.383798211812973,
|
| 157 |
+
"rewards/margins": 0.3239237368106842,
|
| 158 |
+
"rewards/rejected": -0.7077219486236572,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.41841004184100417,
|
| 163 |
+
"grad_norm": 23.30871369910813,
|
| 164 |
+
"learning_rate": 4.821741763807186e-07,
|
| 165 |
+
"logits/chosen": -1.6760666370391846,
|
| 166 |
+
"logits/rejected": -1.5858938694000244,
|
| 167 |
+
"logps/chosen": -309.859375,
|
| 168 |
+
"logps/rejected": -349.28802490234375,
|
| 169 |
+
"loss": 0.5692,
|
| 170 |
+
"rewards/accuracies": 0.6968749761581421,
|
| 171 |
+
"rewards/chosen": -0.44263941049575806,
|
| 172 |
+
"rewards/margins": 0.4911220669746399,
|
| 173 |
+
"rewards/rejected": -0.9337614178657532,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.41841004184100417,
|
| 178 |
+
"eval_logits/chosen": -1.3128046989440918,
|
| 179 |
+
"eval_logits/rejected": -1.1362591981887817,
|
| 180 |
+
"eval_logps/chosen": -299.5389099121094,
|
| 181 |
+
"eval_logps/rejected": -355.61285400390625,
|
| 182 |
+
"eval_loss": 0.5671091079711914,
|
| 183 |
+
"eval_rewards/accuracies": 0.734375,
|
| 184 |
+
"eval_rewards/chosen": -0.3690897822380066,
|
| 185 |
+
"eval_rewards/margins": 0.560417652130127,
|
| 186 |
+
"eval_rewards/rejected": -0.9295073747634888,
|
| 187 |
+
"eval_runtime": 93.5851,
|
| 188 |
+
"eval_samples_per_second": 21.371,
|
| 189 |
+
"eval_steps_per_second": 0.342,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.4602510460251046,
|
| 194 |
+
"grad_norm": 18.679015114468058,
|
| 195 |
+
"learning_rate": 4.747874028753375e-07,
|
| 196 |
+
"logits/chosen": -0.9447735548019409,
|
| 197 |
+
"logits/rejected": -0.7094529867172241,
|
| 198 |
+
"logps/chosen": -332.88055419921875,
|
| 199 |
+
"logps/rejected": -359.24169921875,
|
| 200 |
+
"loss": 0.5669,
|
| 201 |
+
"rewards/accuracies": 0.7281249761581421,
|
| 202 |
+
"rewards/chosen": -0.5535979270935059,
|
| 203 |
+
"rewards/margins": 0.5540369749069214,
|
| 204 |
+
"rewards/rejected": -1.1076349020004272,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.502092050209205,
|
| 209 |
+
"grad_norm": 16.44393963859546,
|
| 210 |
+
"learning_rate": 4.662012913161997e-07,
|
| 211 |
+
"logits/chosen": -0.5595887303352356,
|
| 212 |
+
"logits/rejected": -0.3426797091960907,
|
| 213 |
+
"logps/chosen": -330.4803161621094,
|
| 214 |
+
"logps/rejected": -391.5724792480469,
|
| 215 |
+
"loss": 0.5393,
|
| 216 |
+
"rewards/accuracies": 0.762499988079071,
|
| 217 |
+
"rewards/chosen": -0.7413566708564758,
|
| 218 |
+
"rewards/margins": 0.6902385950088501,
|
| 219 |
+
"rewards/rejected": -1.4315952062606812,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.5439330543933054,
|
| 224 |
+
"grad_norm": 19.802781403351748,
|
| 225 |
+
"learning_rate": 4.5646165232345103e-07,
|
| 226 |
+
"logits/chosen": -0.4343790113925934,
|
| 227 |
+
"logits/rejected": -0.07114946097135544,
|
| 228 |
+
"logps/chosen": -362.0838928222656,
|
| 229 |
+
"logps/rejected": -396.4689025878906,
|
| 230 |
+
"loss": 0.5338,
|
| 231 |
+
"rewards/accuracies": 0.737500011920929,
|
| 232 |
+
"rewards/chosen": -0.6583347916603088,
|
| 233 |
+
"rewards/margins": 0.7389912009239197,
|
| 234 |
+
"rewards/rejected": -1.397326111793518,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.5857740585774058,
|
| 239 |
+
"grad_norm": 21.162587940908836,
|
| 240 |
+
"learning_rate": 4.456204510851956e-07,
|
| 241 |
+
"logits/chosen": -0.3260354697704315,
|
| 242 |
+
"logits/rejected": 0.02392721176147461,
|
| 243 |
+
"logps/chosen": -345.8163757324219,
|
| 244 |
+
"logps/rejected": -386.79254150390625,
|
| 245 |
+
"loss": 0.5444,
|
| 246 |
+
"rewards/accuracies": 0.699999988079071,
|
| 247 |
+
"rewards/chosen": -0.7646363377571106,
|
| 248 |
+
"rewards/margins": 0.6039775013923645,
|
| 249 |
+
"rewards/rejected": -1.368613839149475,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.6276150627615062,
|
| 254 |
+
"grad_norm": 21.73672532038991,
|
| 255 |
+
"learning_rate": 4.337355301007335e-07,
|
| 256 |
+
"logits/chosen": -0.20636191964149475,
|
| 257 |
+
"logits/rejected": 0.28736037015914917,
|
| 258 |
+
"logps/chosen": -364.95477294921875,
|
| 259 |
+
"logps/rejected": -408.7835998535156,
|
| 260 |
+
"loss": 0.5282,
|
| 261 |
+
"rewards/accuracies": 0.715624988079071,
|
| 262 |
+
"rewards/chosen": -0.8889427185058594,
|
| 263 |
+
"rewards/margins": 0.6887326240539551,
|
| 264 |
+
"rewards/rejected": -1.577675223350525,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.6694560669456067,
|
| 269 |
+
"grad_norm": 21.022249543596576,
|
| 270 |
+
"learning_rate": 4.2087030056579986e-07,
|
| 271 |
+
"logits/chosen": 0.2184087336063385,
|
| 272 |
+
"logits/rejected": 0.6583582162857056,
|
| 273 |
+
"logps/chosen": -384.6106872558594,
|
| 274 |
+
"logps/rejected": -409.9974060058594,
|
| 275 |
+
"loss": 0.5064,
|
| 276 |
+
"rewards/accuracies": 0.75,
|
| 277 |
+
"rewards/chosen": -0.9010273218154907,
|
| 278 |
+
"rewards/margins": 0.6821587681770325,
|
| 279 |
+
"rewards/rejected": -1.583186149597168,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.7112970711297071,
|
| 284 |
+
"grad_norm": 19.31934084124254,
|
| 285 |
+
"learning_rate": 4.070934040463998e-07,
|
| 286 |
+
"logits/chosen": 0.3104431927204132,
|
| 287 |
+
"logits/rejected": 0.8817472457885742,
|
| 288 |
+
"logps/chosen": -350.32891845703125,
|
| 289 |
+
"logps/rejected": -423.04266357421875,
|
| 290 |
+
"loss": 0.5153,
|
| 291 |
+
"rewards/accuracies": 0.737500011920929,
|
| 292 |
+
"rewards/chosen": -0.741706132888794,
|
| 293 |
+
"rewards/margins": 0.7905541062355042,
|
| 294 |
+
"rewards/rejected": -1.5322601795196533,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.7531380753138075,
|
| 299 |
+
"grad_norm": 17.210834759479862,
|
| 300 |
+
"learning_rate": 3.9247834624635404e-07,
|
| 301 |
+
"logits/chosen": 1.03719961643219,
|
| 302 |
+
"logits/rejected": 1.7225978374481201,
|
| 303 |
+
"logps/chosen": -386.9837646484375,
|
| 304 |
+
"logps/rejected": -437.9752502441406,
|
| 305 |
+
"loss": 0.5063,
|
| 306 |
+
"rewards/accuracies": 0.75,
|
| 307 |
+
"rewards/chosen": -1.0653451681137085,
|
| 308 |
+
"rewards/margins": 0.8046163320541382,
|
| 309 |
+
"rewards/rejected": -1.8699613809585571,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.7949790794979079,
|
| 314 |
+
"grad_norm": 18.78911912939503,
|
| 315 |
+
"learning_rate": 3.7710310482256523e-07,
|
| 316 |
+
"logits/chosen": 1.121408224105835,
|
| 317 |
+
"logits/rejected": 1.6289663314819336,
|
| 318 |
+
"logps/chosen": -369.95733642578125,
|
| 319 |
+
"logps/rejected": -439.37548828125,
|
| 320 |
+
"loss": 0.5163,
|
| 321 |
+
"rewards/accuracies": 0.746874988079071,
|
| 322 |
+
"rewards/chosen": -1.1254816055297852,
|
| 323 |
+
"rewards/margins": 0.7835719585418701,
|
| 324 |
+
"rewards/rejected": -1.9090534448623657,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.8368200836820083,
|
| 329 |
+
"grad_norm": 18.234200389798175,
|
| 330 |
+
"learning_rate": 3.610497133404795e-07,
|
| 331 |
+
"logits/chosen": 1.011256217956543,
|
| 332 |
+
"logits/rejected": 1.5305925607681274,
|
| 333 |
+
"logps/chosen": -390.50341796875,
|
| 334 |
+
"logps/rejected": -459.53875732421875,
|
| 335 |
+
"loss": 0.5056,
|
| 336 |
+
"rewards/accuracies": 0.7593749761581421,
|
| 337 |
+
"rewards/chosen": -1.0241918563842773,
|
| 338 |
+
"rewards/margins": 0.81651771068573,
|
| 339 |
+
"rewards/rejected": -1.8407094478607178,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.8368200836820083,
|
| 344 |
+
"eval_logits/chosen": 0.9604065418243408,
|
| 345 |
+
"eval_logits/rejected": 1.4953534603118896,
|
| 346 |
+
"eval_logps/chosen": -358.1262512207031,
|
| 347 |
+
"eval_logps/rejected": -447.7286376953125,
|
| 348 |
+
"eval_loss": 0.5104337334632874,
|
| 349 |
+
"eval_rewards/accuracies": 0.78515625,
|
| 350 |
+
"eval_rewards/chosen": -0.9549632668495178,
|
| 351 |
+
"eval_rewards/margins": 0.8957023620605469,
|
| 352 |
+
"eval_rewards/rejected": -1.8506656885147095,
|
| 353 |
+
"eval_runtime": 93.4351,
|
| 354 |
+
"eval_samples_per_second": 21.405,
|
| 355 |
+
"eval_steps_per_second": 0.342,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.8786610878661087,
|
| 360 |
+
"grad_norm": 19.215200164648973,
|
| 361 |
+
"learning_rate": 3.4440382358952115e-07,
|
| 362 |
+
"logits/chosen": 1.0344123840332031,
|
| 363 |
+
"logits/rejected": 1.8179776668548584,
|
| 364 |
+
"logps/chosen": -402.72003173828125,
|
| 365 |
+
"logps/rejected": -460.1229553222656,
|
| 366 |
+
"loss": 0.5066,
|
| 367 |
+
"rewards/accuracies": 0.75,
|
| 368 |
+
"rewards/chosen": -1.0471458435058594,
|
| 369 |
+
"rewards/margins": 0.8377550840377808,
|
| 370 |
+
"rewards/rejected": -1.884900689125061,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.9205020920502092,
|
| 375 |
+
"grad_norm": 22.998405474723544,
|
| 376 |
+
"learning_rate": 3.272542485937368e-07,
|
| 377 |
+
"logits/chosen": 0.9624277353286743,
|
| 378 |
+
"logits/rejected": 1.7323452234268188,
|
| 379 |
+
"logps/chosen": -374.59521484375,
|
| 380 |
+
"logps/rejected": -442.35650634765625,
|
| 381 |
+
"loss": 0.4894,
|
| 382 |
+
"rewards/accuracies": 0.784375011920929,
|
| 383 |
+
"rewards/chosen": -0.9290931820869446,
|
| 384 |
+
"rewards/margins": 0.8601258397102356,
|
| 385 |
+
"rewards/rejected": -1.7892189025878906,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.9623430962343096,
|
| 390 |
+
"grad_norm": 17.30285752760936,
|
| 391 |
+
"learning_rate": 3.096924887558854e-07,
|
| 392 |
+
"logits/chosen": 0.9907492399215698,
|
| 393 |
+
"logits/rejected": 1.7148042917251587,
|
| 394 |
+
"logps/chosen": -386.2179260253906,
|
| 395 |
+
"logps/rejected": -461.5531311035156,
|
| 396 |
+
"loss": 0.5016,
|
| 397 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 398 |
+
"rewards/chosen": -1.0336068868637085,
|
| 399 |
+
"rewards/margins": 0.8331363797187805,
|
| 400 |
+
"rewards/rejected": -1.8667433261871338,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 1.00418410041841,
|
| 405 |
+
"grad_norm": 17.247084086602182,
|
| 406 |
+
"learning_rate": 2.9181224366319943e-07,
|
| 407 |
+
"logits/chosen": 0.6638993620872498,
|
| 408 |
+
"logits/rejected": 1.3943182229995728,
|
| 409 |
+
"logps/chosen": -376.9907531738281,
|
| 410 |
+
"logps/rejected": -443.1883850097656,
|
| 411 |
+
"loss": 0.4867,
|
| 412 |
+
"rewards/accuracies": 0.734375,
|
| 413 |
+
"rewards/chosen": -0.9322942495346069,
|
| 414 |
+
"rewards/margins": 0.7798601388931274,
|
| 415 |
+
"rewards/rejected": -1.7121543884277344,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 1.0460251046025104,
|
| 420 |
+
"grad_norm": 27.106663640280367,
|
| 421 |
+
"learning_rate": 2.7370891215954565e-07,
|
| 422 |
+
"logits/chosen": 1.1816644668579102,
|
| 423 |
+
"logits/rejected": 2.409132480621338,
|
| 424 |
+
"logps/chosen": -390.9169006347656,
|
| 425 |
+
"logps/rejected": -476.27032470703125,
|
| 426 |
+
"loss": 0.3951,
|
| 427 |
+
"rewards/accuracies": 0.796875,
|
| 428 |
+
"rewards/chosen": -1.0209319591522217,
|
| 429 |
+
"rewards/margins": 1.2181396484375,
|
| 430 |
+
"rewards/rejected": -2.2390713691711426,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 1.0878661087866108,
|
| 435 |
+
"grad_norm": 26.07881941521921,
|
| 436 |
+
"learning_rate": 2.55479083351317e-07,
|
| 437 |
+
"logits/chosen": 1.634642243385315,
|
| 438 |
+
"logits/rejected": 2.725491762161255,
|
| 439 |
+
"logps/chosen": -382.5552673339844,
|
| 440 |
+
"logps/rejected": -490.21063232421875,
|
| 441 |
+
"loss": 0.3905,
|
| 442 |
+
"rewards/accuracies": 0.815625011920929,
|
| 443 |
+
"rewards/chosen": -1.040998935699463,
|
| 444 |
+
"rewards/margins": 1.2747693061828613,
|
| 445 |
+
"rewards/rejected": -2.3157684803009033,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 1.1297071129707112,
|
| 450 |
+
"grad_norm": 25.1217443727487,
|
| 451 |
+
"learning_rate": 2.3722002126275822e-07,
|
| 452 |
+
"logits/chosen": 1.4133918285369873,
|
| 453 |
+
"logits/rejected": 2.5140790939331055,
|
| 454 |
+
"logps/chosen": -383.543212890625,
|
| 455 |
+
"logps/rejected": -506.23443603515625,
|
| 456 |
+
"loss": 0.3874,
|
| 457 |
+
"rewards/accuracies": 0.8343750238418579,
|
| 458 |
+
"rewards/chosen": -1.2637007236480713,
|
| 459 |
+
"rewards/margins": 1.3295375108718872,
|
| 460 |
+
"rewards/rejected": -2.593238353729248,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 1.1715481171548117,
|
| 465 |
+
"grad_norm": 25.52463427219616,
|
| 466 |
+
"learning_rate": 2.19029145890313e-07,
|
| 467 |
+
"logits/chosen": 1.2717490196228027,
|
| 468 |
+
"logits/rejected": 2.5123631954193115,
|
| 469 |
+
"logps/chosen": -398.04620361328125,
|
| 470 |
+
"logps/rejected": -503.71759033203125,
|
| 471 |
+
"loss": 0.3954,
|
| 472 |
+
"rewards/accuracies": 0.831250011920929,
|
| 473 |
+
"rewards/chosen": -1.2178623676300049,
|
| 474 |
+
"rewards/margins": 1.3320947885513306,
|
| 475 |
+
"rewards/rejected": -2.549957275390625,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 1.213389121338912,
|
| 480 |
+
"grad_norm": 19.330696144971178,
|
| 481 |
+
"learning_rate": 2.0100351342479216e-07,
|
| 482 |
+
"logits/chosen": 1.1333314180374146,
|
| 483 |
+
"logits/rejected": 2.094285011291504,
|
| 484 |
+
"logps/chosen": -396.33905029296875,
|
| 485 |
+
"logps/rejected": -513.3264770507812,
|
| 486 |
+
"loss": 0.381,
|
| 487 |
+
"rewards/accuracies": 0.84375,
|
| 488 |
+
"rewards/chosen": -1.1736336946487427,
|
| 489 |
+
"rewards/margins": 1.3145819902420044,
|
| 490 |
+
"rewards/rejected": -2.488215684890747,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 1.2552301255230125,
|
| 495 |
+
"grad_norm": 25.82255359759937,
|
| 496 |
+
"learning_rate": 1.8323929841460178e-07,
|
| 497 |
+
"logits/chosen": 0.8372614979743958,
|
| 498 |
+
"logits/rejected": 1.8510090112686157,
|
| 499 |
+
"logps/chosen": -397.23175048828125,
|
| 500 |
+
"logps/rejected": -493.86578369140625,
|
| 501 |
+
"loss": 0.3744,
|
| 502 |
+
"rewards/accuracies": 0.828125,
|
| 503 |
+
"rewards/chosen": -1.2290160655975342,
|
| 504 |
+
"rewards/margins": 1.2341445684432983,
|
| 505 |
+
"rewards/rejected": -2.463160753250122,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 1.2552301255230125,
|
| 510 |
+
"eval_logits/chosen": 0.9870985150337219,
|
| 511 |
+
"eval_logits/rejected": 1.8526870012283325,
|
| 512 |
+
"eval_logps/chosen": -397.17913818359375,
|
| 513 |
+
"eval_logps/rejected": -518.840576171875,
|
| 514 |
+
"eval_loss": 0.5120064616203308,
|
| 515 |
+
"eval_rewards/accuracies": 0.76171875,
|
| 516 |
+
"eval_rewards/chosen": -1.3454917669296265,
|
| 517 |
+
"eval_rewards/margins": 1.2162930965423584,
|
| 518 |
+
"eval_rewards/rejected": -2.5617849826812744,
|
| 519 |
+
"eval_runtime": 93.0364,
|
| 520 |
+
"eval_samples_per_second": 21.497,
|
| 521 |
+
"eval_steps_per_second": 0.344,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 1.297071129707113,
|
| 526 |
+
"grad_norm": 40.35910050212018,
|
| 527 |
+
"learning_rate": 1.6583128063291573e-07,
|
| 528 |
+
"logits/chosen": 0.8788517713546753,
|
| 529 |
+
"logits/rejected": 2.132133722305298,
|
| 530 |
+
"logps/chosen": -439.266845703125,
|
| 531 |
+
"logps/rejected": -540.0487060546875,
|
| 532 |
+
"loss": 0.3858,
|
| 533 |
+
"rewards/accuracies": 0.8125,
|
| 534 |
+
"rewards/chosen": -1.3913695812225342,
|
| 535 |
+
"rewards/margins": 1.3974393606185913,
|
| 536 |
+
"rewards/rejected": -2.788808822631836,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 1.3389121338912133,
|
| 541 |
+
"grad_norm": 20.432466646869965,
|
| 542 |
+
"learning_rate": 1.488723393865766e-07,
|
| 543 |
+
"logits/chosen": 0.7624516487121582,
|
| 544 |
+
"logits/rejected": 1.9204070568084717,
|
| 545 |
+
"logps/chosen": -394.59002685546875,
|
| 546 |
+
"logps/rejected": -506.1437072753906,
|
| 547 |
+
"loss": 0.3764,
|
| 548 |
+
"rewards/accuracies": 0.859375,
|
| 549 |
+
"rewards/chosen": -1.1876527070999146,
|
| 550 |
+
"rewards/margins": 1.364680528640747,
|
| 551 |
+
"rewards/rejected": -2.552333354949951,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 1.3807531380753137,
|
| 556 |
+
"grad_norm": 24.117995597732122,
|
| 557 |
+
"learning_rate": 1.3245295796480788e-07,
|
| 558 |
+
"logits/chosen": 1.2566739320755005,
|
| 559 |
+
"logits/rejected": 2.4089713096618652,
|
| 560 |
+
"logps/chosen": -403.59539794921875,
|
| 561 |
+
"logps/rejected": -535.1661987304688,
|
| 562 |
+
"loss": 0.3785,
|
| 563 |
+
"rewards/accuracies": 0.8656250238418579,
|
| 564 |
+
"rewards/chosen": -1.3696684837341309,
|
| 565 |
+
"rewards/margins": 1.466200828552246,
|
| 566 |
+
"rewards/rejected": -2.835869312286377,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 1.4225941422594142,
|
| 571 |
+
"grad_norm": 26.170896618734517,
|
| 572 |
+
"learning_rate": 1.1666074087171627e-07,
|
| 573 |
+
"logits/chosen": 1.6827380657196045,
|
| 574 |
+
"logits/rejected": 2.633002519607544,
|
| 575 |
+
"logps/chosen": -409.3313903808594,
|
| 576 |
+
"logps/rejected": -560.8604736328125,
|
| 577 |
+
"loss": 0.3673,
|
| 578 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 579 |
+
"rewards/chosen": -1.4851205348968506,
|
| 580 |
+
"rewards/margins": 1.4507629871368408,
|
| 581 |
+
"rewards/rejected": -2.9358837604522705,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 1.4644351464435146,
|
| 586 |
+
"grad_norm": 20.814302126974475,
|
| 587 |
+
"learning_rate": 1.0157994641835734e-07,
|
| 588 |
+
"logits/chosen": 1.3171794414520264,
|
| 589 |
+
"logits/rejected": 2.3436312675476074,
|
| 590 |
+
"logps/chosen": -415.7732849121094,
|
| 591 |
+
"logps/rejected": -539.7247924804688,
|
| 592 |
+
"loss": 0.3782,
|
| 593 |
+
"rewards/accuracies": 0.8218749761581421,
|
| 594 |
+
"rewards/chosen": -1.373610496520996,
|
| 595 |
+
"rewards/margins": 1.4075970649719238,
|
| 596 |
+
"rewards/rejected": -2.78120756149292,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 1.506276150627615,
|
| 601 |
+
"grad_norm": 21.900774053482067,
|
| 602 |
+
"learning_rate": 8.729103716819111e-08,
|
| 603 |
+
"logits/chosen": 1.4213756322860718,
|
| 604 |
+
"logits/rejected": 2.515977382659912,
|
| 605 |
+
"logps/chosen": -440.63311767578125,
|
| 606 |
+
"logps/rejected": -566.4862060546875,
|
| 607 |
+
"loss": 0.3606,
|
| 608 |
+
"rewards/accuracies": 0.846875011920929,
|
| 609 |
+
"rewards/chosen": -1.460523009300232,
|
| 610 |
+
"rewards/margins": 1.5627386569976807,
|
| 611 |
+
"rewards/rejected": -3.023261308670044,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 1.5481171548117154,
|
| 616 |
+
"grad_norm": 24.209127764578607,
|
| 617 |
+
"learning_rate": 7.387025063449081e-08,
|
| 618 |
+
"logits/chosen": 1.1703312397003174,
|
| 619 |
+
"logits/rejected": 2.1595053672790527,
|
| 620 |
+
"logps/chosen": -427.52734375,
|
| 621 |
+
"logps/rejected": -536.962646484375,
|
| 622 |
+
"loss": 0.3769,
|
| 623 |
+
"rewards/accuracies": 0.824999988079071,
|
| 624 |
+
"rewards/chosen": -1.4095176458358765,
|
| 625 |
+
"rewards/margins": 1.3942286968231201,
|
| 626 |
+
"rewards/rejected": -2.803746461868286,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 1.5899581589958158,
|
| 631 |
+
"grad_norm": 25.431449474931732,
|
| 632 |
+
"learning_rate": 6.138919252022435e-08,
|
| 633 |
+
"logits/chosen": 1.452207326889038,
|
| 634 |
+
"logits/rejected": 2.379581928253174,
|
| 635 |
+
"logps/chosen": -434.0857849121094,
|
| 636 |
+
"logps/rejected": -585.3619384765625,
|
| 637 |
+
"loss": 0.366,
|
| 638 |
+
"rewards/accuracies": 0.84375,
|
| 639 |
+
"rewards/chosen": -1.3957841396331787,
|
| 640 |
+
"rewards/margins": 1.5939209461212158,
|
| 641 |
+
"rewards/rejected": -2.9897048473358154,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 1.6317991631799162,
|
| 646 |
+
"grad_norm": 22.87966472658563,
|
| 647 |
+
"learning_rate": 4.991445467064689e-08,
|
| 648 |
+
"logits/chosen": 1.3533505201339722,
|
| 649 |
+
"logits/rejected": 2.646073579788208,
|
| 650 |
+
"logps/chosen": -424.43267822265625,
|
| 651 |
+
"logps/rejected": -543.8640747070312,
|
| 652 |
+
"loss": 0.3723,
|
| 653 |
+
"rewards/accuracies": 0.84375,
|
| 654 |
+
"rewards/chosen": -1.4647841453552246,
|
| 655 |
+
"rewards/margins": 1.4622496366500854,
|
| 656 |
+
"rewards/rejected": -2.9270341396331787,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 1.6736401673640167,
|
| 661 |
+
"grad_norm": 21.87896636487824,
|
| 662 |
+
"learning_rate": 3.9507259776993954e-08,
|
| 663 |
+
"logits/chosen": 1.3778632879257202,
|
| 664 |
+
"logits/rejected": 2.785320281982422,
|
| 665 |
+
"logps/chosen": -429.8521423339844,
|
| 666 |
+
"logps/rejected": -541.7579345703125,
|
| 667 |
+
"loss": 0.351,
|
| 668 |
+
"rewards/accuracies": 0.856249988079071,
|
| 669 |
+
"rewards/chosen": -1.3404476642608643,
|
| 670 |
+
"rewards/margins": 1.5822999477386475,
|
| 671 |
+
"rewards/rejected": -2.922747850418091,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 1.6736401673640167,
|
| 676 |
+
"eval_logits/chosen": 1.4005358219146729,
|
| 677 |
+
"eval_logits/rejected": 2.178215265274048,
|
| 678 |
+
"eval_logps/chosen": -414.581298828125,
|
| 679 |
+
"eval_logps/rejected": -536.5388793945312,
|
| 680 |
+
"eval_loss": 0.5018947124481201,
|
| 681 |
+
"eval_rewards/accuracies": 0.76953125,
|
| 682 |
+
"eval_rewards/chosen": -1.519513487815857,
|
| 683 |
+
"eval_rewards/margins": 1.219254732131958,
|
| 684 |
+
"eval_rewards/rejected": -2.7387683391571045,
|
| 685 |
+
"eval_runtime": 92.8387,
|
| 686 |
+
"eval_samples_per_second": 21.543,
|
| 687 |
+
"eval_steps_per_second": 0.345,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 1.715481171548117,
|
| 692 |
+
"grad_norm": 25.54848296222211,
|
| 693 |
+
"learning_rate": 3.022313472693447e-08,
|
| 694 |
+
"logits/chosen": 1.3364863395690918,
|
| 695 |
+
"logits/rejected": 2.4736454486846924,
|
| 696 |
+
"logps/chosen": -421.3280334472656,
|
| 697 |
+
"logps/rejected": -570.7842407226562,
|
| 698 |
+
"loss": 0.375,
|
| 699 |
+
"rewards/accuracies": 0.846875011920929,
|
| 700 |
+
"rewards/chosen": -1.4803876876831055,
|
| 701 |
+
"rewards/margins": 1.5594526529312134,
|
| 702 |
+
"rewards/rejected": -3.0398402214050293,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 1.7573221757322175,
|
| 707 |
+
"grad_norm": 30.706537474329608,
|
| 708 |
+
"learning_rate": 2.2111614344599684e-08,
|
| 709 |
+
"logits/chosen": 1.4454032182693481,
|
| 710 |
+
"logits/rejected": 2.485903263092041,
|
| 711 |
+
"logps/chosen": -436.48260498046875,
|
| 712 |
+
"logps/rejected": -557.8076171875,
|
| 713 |
+
"loss": 0.361,
|
| 714 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 715 |
+
"rewards/chosen": -1.5660855770111084,
|
| 716 |
+
"rewards/margins": 1.5376800298690796,
|
| 717 |
+
"rewards/rejected": -3.1037657260894775,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 1.799163179916318,
|
| 722 |
+
"grad_norm": 23.906131529756504,
|
| 723 |
+
"learning_rate": 1.521597710086439e-08,
|
| 724 |
+
"logits/chosen": 1.5828850269317627,
|
| 725 |
+
"logits/rejected": 2.5288591384887695,
|
| 726 |
+
"logps/chosen": -428.4192810058594,
|
| 727 |
+
"logps/rejected": -542.4400634765625,
|
| 728 |
+
"loss": 0.369,
|
| 729 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 730 |
+
"rewards/chosen": -1.4776898622512817,
|
| 731 |
+
"rewards/margins": 1.4163352251052856,
|
| 732 |
+
"rewards/rejected": -2.8940250873565674,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 1.8410041841004183,
|
| 737 |
+
"grad_norm": 28.22708684609244,
|
| 738 |
+
"learning_rate": 9.57301420397924e-09,
|
| 739 |
+
"logits/chosen": 1.6976782083511353,
|
| 740 |
+
"logits/rejected": 2.871784210205078,
|
| 741 |
+
"logps/chosen": -424.5814514160156,
|
| 742 |
+
"logps/rejected": -530.4151000976562,
|
| 743 |
+
"loss": 0.3655,
|
| 744 |
+
"rewards/accuracies": 0.8531249761581421,
|
| 745 |
+
"rewards/chosen": -1.5554591417312622,
|
| 746 |
+
"rewards/margins": 1.416295051574707,
|
| 747 |
+
"rewards/rejected": -2.971754550933838,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 1.8828451882845187,
|
| 752 |
+
"grad_norm": 21.46648253004346,
|
| 753 |
+
"learning_rate": 5.212833302556258e-09,
|
| 754 |
+
"logits/chosen": 1.5777668952941895,
|
| 755 |
+
"logits/rejected": 2.6181626319885254,
|
| 756 |
+
"logps/chosen": -431.70086669921875,
|
| 757 |
+
"logps/rejected": -552.4868774414062,
|
| 758 |
+
"loss": 0.3454,
|
| 759 |
+
"rewards/accuracies": 0.871874988079071,
|
| 760 |
+
"rewards/chosen": -1.5065062046051025,
|
| 761 |
+
"rewards/margins": 1.5084437131881714,
|
| 762 |
+
"rewards/rejected": -3.0149497985839844,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 1.9246861924686192,
|
| 767 |
+
"grad_norm": 24.44249151761102,
|
| 768 |
+
"learning_rate": 2.158697848236607e-09,
|
| 769 |
+
"logits/chosen": 1.7532329559326172,
|
| 770 |
+
"logits/rejected": 2.691193103790283,
|
| 771 |
+
"logps/chosen": -401.5673828125,
|
| 772 |
+
"logps/rejected": -556.6257934570312,
|
| 773 |
+
"loss": 0.3685,
|
| 774 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 775 |
+
"rewards/chosen": -1.477992057800293,
|
| 776 |
+
"rewards/margins": 1.5126928091049194,
|
| 777 |
+
"rewards/rejected": -2.990684986114502,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 1.9665271966527196,
|
| 782 |
+
"grad_norm": 25.733721301575518,
|
| 783 |
+
"learning_rate": 4.269029751107489e-10,
|
| 784 |
+
"logits/chosen": 1.2774773836135864,
|
| 785 |
+
"logits/rejected": 2.5834484100341797,
|
| 786 |
+
"logps/chosen": -431.2181701660156,
|
| 787 |
+
"logps/rejected": -525.5928955078125,
|
| 788 |
+
"loss": 0.3641,
|
| 789 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 790 |
+
"rewards/chosen": -1.4548835754394531,
|
| 791 |
+
"rewards/margins": 1.3838748931884766,
|
| 792 |
+
"rewards/rejected": -2.838758707046509,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 2.0,
|
| 797 |
+
"step": 478,
|
| 798 |
+
"total_flos": 0.0,
|
| 799 |
+
"train_loss": 0.47010813497599196,
|
| 800 |
+
"train_runtime": 14181.5338,
|
| 801 |
+
"train_samples_per_second": 8.622,
|
| 802 |
+
"train_steps_per_second": 0.034
|
| 803 |
+
}
|
| 804 |
+
],
|
| 805 |
+
"logging_steps": 10,
|
| 806 |
+
"max_steps": 478,
|
| 807 |
+
"num_input_tokens_seen": 0,
|
| 808 |
+
"num_train_epochs": 2,
|
| 809 |
+
"save_steps": 100,
|
| 810 |
+
"stateful_callbacks": {
|
| 811 |
+
"TrainerControl": {
|
| 812 |
+
"args": {
|
| 813 |
+
"should_epoch_stop": false,
|
| 814 |
+
"should_evaluate": false,
|
| 815 |
+
"should_log": false,
|
| 816 |
+
"should_save": true,
|
| 817 |
+
"should_training_stop": true
|
| 818 |
+
},
|
| 819 |
+
"attributes": {}
|
| 820 |
+
}
|
| 821 |
+
},
|
| 822 |
+
"total_flos": 0.0,
|
| 823 |
+
"train_batch_size": 8,
|
| 824 |
+
"trial_name": null,
|
| 825 |
+
"trial_params": null
|
| 826 |
+
}
|